Summary

The PDZ-GEF protein Dizzy (Dzy) and its downstream GTPase Rap1 have pleiotropic roles during development of the Drosophila embryo. Here, we show that maternally provided Dzy and Rap1 first function during ventral furrow formation (VFF) where they are critical to guarantee rapid apical cell constrictions. Contraction of the apical actomyosin filament system occurs independently of Dzy and Rap1, but loss of Dzy results in a delayed establishment of the apical adherens junction (AJ) belt, whereas in the absence of Rap1 only a fragmentary apical AJ belt is formed in the epithelium. The timely establishment of apical AJs appears to be essential for coupling actomyosin contractions to cell shape change and to assure completion of the ventral furrow. Immediately after VFF, the downregulation of Dzy and Rap1 is necessary to allow normal mesodermal development to continue after the epithelial-to-mesenchymal transition, as overexpression of Dzy or of constitutively active Rap1 compromises mesodermal migration and monolayer formation. We propose that Dzy and Rap1 are crucial factors regulating the dynamics of AJs during gastrulation.

Introduction

Cell shape changes are essential during development as principal events of morphogenesis. In epithelia, they require an actomyosin cytoskeleton that is properly connected to the cell membranes via adherens junctions (AJs). AJs consist of clusters of the transmembrane protein DE-Cadherin (DE-Cad) that mediates adhesion to the neighbouring cell via extracellular homophilic interactions. On the cytoplasmic side, DE-Cad is complexed with different catenins to connect to the cytoskeleton in a dynamic and not yet fully characterized mechanism (reviewed in Gates and Peifer, 2005; Harris et al., 2009). In Drosophila, AJs are set up during cellularization (reviewed in Tepass et al., 2001) and are critical prerequisites for the rapid cell shape changes and epithelial rearrangements taking place at gastrulation during the formation of the ventral furrow (Dawes-Hoang et al., 2005; Kölsch et al., 2007).

Ventral furrow formation (VFF) in Drosophila is an attractive model for analyzing the junctional mechanics in epithelia as it is easy to access and amenable to genetic manipulation (reviewed in Leptin, 2005). The cell shape changes that are observed during VFF include flattening of the apical cortex in a band of mid-ventral cells, followed by apical constriction (visible as an indentation of ventral tissue in cross-sections) and invagination into the interior of the embryo (Leptin and Grunewald, 1990; Sweeton et al., 1991). These processes are thought to be brought about by contraction of an apical actomyosin meshwork that is connected to apically localized AJs to convey contractive forces and to achieve cell shape change (Dawes-Hoang et al., 2005; Haruta et al., 2010; Kölsch et al., 2007; Martin et al., 2009; Sawyer et al., 2009). Parallel to these events in mid-ventral cells, latero-ventral cells become stretched and move towards the midline to close the furrow (Leptin and Grunewald, 1990).

A signalling cascade has been identified that induces the cell shape changes which characterize the ventral furrow. Two transcriptional targets of the mesodermal determinant Twist (Twi), Folded gastrulation (Fog) and T48, ensure the apical accumulation of the G-nucleotide exchange factor RhoGEF2 in a parallel fashion (Kölsch et al., 2007). The apically concentrated RhoGEF2 is thought to activate the small GTPase Rho (Barrett et al., 1997; Häcker and Perrimon, 1998) which in turn activates Myosin II (MyoII) via Rho kinase (Nikolaidou and Barrett, 2004). This activation of MyoII leads to the apical assembly of contracting actomyosin which, while being tethered to apical AJs, constricts the apices of mid-ventral cells (Dawes-Hoang et al., 2005; Martin et al., 2009; Sawyer et al., 2009). It is unclear, however, what ensures the allocation of the apical AJs at the onset of gastrulation.

Here, we analyze the role of the PDZ-GEF protein Dizzy (Dzy) and its downstream small GTPase Rap1 during VFF. Dzy is highly conserved among bilaterians and has been recognized as a GEF for Rap1 in several model systems (de Rooij et al., 1999; Kawajiri et al., 2000). In Drosophila, it is involved in wing, eye and ovary development and is required for stem cell maintenance in spermatogenesis, for macrophage migration and dorsal closure. In most cases it regulates cell adhesion that is mediated by DE-Cad or integrins (Boettner and Van Aelst, 2007; Huelsmann et al., 2006; Lee et al., 2002; Wang et al., 2006). In this work, we show that Dzy has a role in VFF, since lack of maternally provided Dzy results in a slowdown of apical constriction in ventral cells, probably caused by a slowdown of apical AJ assembly occurring around the epithelium.

Rap1 also has numerous functions during development throughout various model systems most of which are linked to regulation of adhesion to neighbouring cells or the extracellular matrix (reviewed in Boettner and Van Aelst, 2009). In particular, it has been shown that loss of rap1 leads to defects in cell–cell adhesion, due to disturbed distribution of AJs in cell membranes (Knox and Brown, 2002). In addition to other processes involving Rap1 function, the formation of the ventral furrow has been known to depend on maternally provided Rap1 (Asha et al., 1999). Recently, this role has been characterized further: when maternal contribution of Rap1 is lost, apical constriction of ventral cells remains incomplete as the contracting actomyosin is not properly attached to cell membranes (Sawyer et al., 2009). Here, we extend these studies and find the formation of apical AJs to be severely compromised in the absence of maternal Rap1. The major AJ component DE-Cadherin (DE-Cad) is aberrantly dispersed within cells of the entire epithelium and is not accumulated into an apical belt. In addition, we show that Dzy and Rap1 cause defects when ectopically expressed in the internalized mesoderm. After epithelial-to-mesenchymal transition (EMT) mesodermal cells are impeded in their migration and monolayer formation.

Results

Ventral furrow formation requires maternal contribution of Dzy and Rap1

We generated germline clones (GLC), eliminating the maternal contribution of Dzy or Rap1, to address their early function in the Drosophila embryo. We found that ventral furrow formation (VFF) is the first process to be dependent on Dzy as it is severely disturbed in the GLC (Fig. 1). In wild-type embryos the ventral furrow forms during stage 6 (Fig. 1D,D′) and has closed at stage 7 (Fig. 1G,G′). In contrast, in dzy GLC the furrow is rudimentary at stage 6 (Fig. 1E,E′) and is closed incompletely at stage 7 (Fig. 1H,H′).

Fig. 1.

Maternally provided Dzy and Rap1 are required for proper VFF. Wild-type embryos (A,A′,D,D′,G,G′), dzy GLC (B,B′,E,E′,H,H′) and rap1 GLC (C,C′,F,F′,I,I′) stained for Twi (brown) and Eve protein (black). (AC′) Cellularization is not affected in dzy or rap1 GLC. (DF′) During stage 6a, a deep ventral furrow forms in wild-type embryos (D,D′), but only a shallow groove forms in dzy GLC (E,E′). In rap1 GLC, a furrow is not evident (F,F′). (GI′) At stage 7 the mesoderm is fully invaginated in wild-type embryos (G,G′), whereas the furrow is only partially closed in dzy GLC (H,H′, arrowhead). In rap1 GLC, the ventral furrow has failed to form (I,I′). Anterior left; A–I, sagittal view; A′–I′, ventral view (same embryos).

Fig. 1.

Maternally provided Dzy and Rap1 are required for proper VFF. Wild-type embryos (A,A′,D,D′,G,G′), dzy GLC (B,B′,E,E′,H,H′) and rap1 GLC (C,C′,F,F′,I,I′) stained for Twi (brown) and Eve protein (black). (AC′) Cellularization is not affected in dzy or rap1 GLC. (DF′) During stage 6a, a deep ventral furrow forms in wild-type embryos (D,D′), but only a shallow groove forms in dzy GLC (E,E′). In rap1 GLC, a furrow is not evident (F,F′). (GI′) At stage 7 the mesoderm is fully invaginated in wild-type embryos (G,G′), whereas the furrow is only partially closed in dzy GLC (H,H′, arrowhead). In rap1 GLC, the ventral furrow has failed to form (I,I′). Anterior left; A–I, sagittal view; A′–I′, ventral view (same embryos).

Dzy activates the small GTPase Rap1 in other morphogenetic events of Drosophila (Boettner and Van Aelst, 2007; Huelsmann et al., 2006). Since Rap1 has also been implicated in VFF (Asha et al., 1999; Sawyer et al., 2009), the phenotypes of rap1 GLC were compared to dzy GLC. VFF is more severely affected in rap1 GLC: at stage 6 no tissue indentation is visible at the ventral embryo surface (Fig. 1F,F′) and no furrow has formed at stage 7 (Fig. 1I,I′), leaving the mesoderm primordium not internalized. These VFF phenotypes of dzy or rap1 GLC are independent of the zygotic genotype of the embryo. They are seen in embryos lacking both maternal and zygotic dzy or rap1 contribution and in embryos merely lacking the maternal contribution.

The failure to form a proper ventral furrow is not due to defects in pattern formation in dzy or rap1 GLC. The mesodermal transcription factors Twi (Fig. 1) and Sna (supplementary material Fig. S1G–I) and the dorsal determinant pnr (supplementary material Fig. S1A–C) are expressed in their normal dorsoventral domains during and after cellularization in dzy and rap1 GLC. Hence, loss of Dzy or Rap1 does not interfere with dorsoventral patterning. Consequently, some aspect of the cellular machinery that executes the shape changes during VFF depends on Dzy and Rap1.

Terminal patterning is not affected in dzy and rap1 GLC, either. For instance, hkb RNA is expressed at the anterior and posterior pole just like in wild-type (supplementary material Fig. S1D–F), in contrast to earlier reports stating that hkb mRNA is reduced in rap1 GLC (Mishra et al., 2005). The posterior midgut (PMG) invagination and the germ band extension commence in dzy or rap1 GLC like in wild-type (Fig. 1; supplementary material Movie 1). This can easily be followed via movement of the pole cells as they are anchored to the epithelium of the PMG primordium. In wild-type and in dzy and rap1 GLC the pole cells initially move dorsally with essentially the same speed (supplementary material Movie 1).

For further analysis of the morphogenetic events of VFF in wild-type and mutant embryos we introduced a fine staging of stage 6 (supplementary material Fig. S2): 6e (early), 6m (middle) and 6l (late). This staging is based on the movement of the pole cells indicating the progress of PMG invagination which is largely synchronous to the progress of VFF in wild-type and not affected in dzy and rap1 GLC.

Apical constriction of ventral cells is compromised in dzy and rap1 germline clones

In wild-type, VFF is thought to be brought about by fast and extensive shape changes of ventral cells. To analyze possible alterations of these shape changes in dzy and rap1 GLC, cell membranes were marked in cross sections of fixed embryos (Fig. 2A–L, respective left panels) or in surface views of live embryos (Fig. 2A–L, respective right panels). In wild-type, immediately after cellularization is completed, mid-ventral cells flatten apically and begin to constrict their apices. The apical constrictions are enhanced and the cells shorten in the apicobasal direction, thereby forming a tissue indentation (Fig. 2B). The indentation deepens (Fig. 2C) so that an epithelial tube is closed and fully internalized during stage 7 (Fig. 2D). In dzy GLC, mid-ventral cells flatten at stage 6m, but do not shorten leading to a high columnar epithelium (Fig. 2F). Apices are less constricted than in wild-type, the furrow is shallow at stage 6l (Fig. 2G) and not completely closed at stage 7 (Fig. 2H).

Fig. 2.

Maternal loss of dzy or rap1 affects constriction of ventral cells. (AL) Cross sections stained for Neurotactin (left) and stills of confocal live-imaging using Spider:GFP (right) of wild-type embryos (A–D), dzy GLC (E–H) and rap1 GLC (I–L). Wild-type: After completion of cellularization (A), cells from a mid-ventral band constrict their apices and the prospective mesoderm starts to invaginate, visible as an indentation in cross-sections (B, arrowhead). The indentation deepens (C) and the mesoderm has been entirely internalized into the ventral furrow by stage 7 (D). dzy: Cellularization proceeds normally (E), but cell constriction is severely delayed (F) leading to a slow-forming and non-persistent ventral furrow (G,H). rap1: After normal cellularization (I) apical constriction occurs in a disperse pattern resulting in a random arrangement of constricted and unconstricted cells (I–L, blue and red arrowheads, resp.), but the ventral furrow never forms. Cross-sections: ventral down, at 50% egg length. Surface views taken at 5 µm depth. (MO) Plots of apical area (red) and cell eccentricity (blue) over the first 20 minutes of gastrulation from five mid-ventral cells. In the wild-type, plots stop at 10 minutes when mid-ventral cells have been internalized. (PR) Histograms showing the time-dependent distribution of constriction levels among mid-ventral cells, coded by colour. Constriction levels indicate the amount of apical area decrease relative to time point 0 (100% constriction  =  reduction to zero area; 0% constriction  =  no change in area; <0%  =  area enlargement). In wild-type embryos, nearly all mid-ventral cells have reached strong constriction levels after 10 minutes (blue fractions: 60–100% constriction). In dzy GLC, the ventral epithelium needs about twice as long to achieve an equal fraction of constricted cells. In rap1 GLC, unconstricted or even bloated cells remain throughout gastrulation indicated by the red fractions (WT: n = 39; dzy: n = 30; rap1: n = 31).

Fig. 2.

Maternal loss of dzy or rap1 affects constriction of ventral cells. (AL) Cross sections stained for Neurotactin (left) and stills of confocal live-imaging using Spider:GFP (right) of wild-type embryos (A–D), dzy GLC (E–H) and rap1 GLC (I–L). Wild-type: After completion of cellularization (A), cells from a mid-ventral band constrict their apices and the prospective mesoderm starts to invaginate, visible as an indentation in cross-sections (B, arrowhead). The indentation deepens (C) and the mesoderm has been entirely internalized into the ventral furrow by stage 7 (D). dzy: Cellularization proceeds normally (E), but cell constriction is severely delayed (F) leading to a slow-forming and non-persistent ventral furrow (G,H). rap1: After normal cellularization (I) apical constriction occurs in a disperse pattern resulting in a random arrangement of constricted and unconstricted cells (I–L, blue and red arrowheads, resp.), but the ventral furrow never forms. Cross-sections: ventral down, at 50% egg length. Surface views taken at 5 µm depth. (MO) Plots of apical area (red) and cell eccentricity (blue) over the first 20 minutes of gastrulation from five mid-ventral cells. In the wild-type, plots stop at 10 minutes when mid-ventral cells have been internalized. (PR) Histograms showing the time-dependent distribution of constriction levels among mid-ventral cells, coded by colour. Constriction levels indicate the amount of apical area decrease relative to time point 0 (100% constriction  =  reduction to zero area; 0% constriction  =  no change in area; <0%  =  area enlargement). In wild-type embryos, nearly all mid-ventral cells have reached strong constriction levels after 10 minutes (blue fractions: 60–100% constriction). In dzy GLC, the ventral epithelium needs about twice as long to achieve an equal fraction of constricted cells. In rap1 GLC, unconstricted or even bloated cells remain throughout gastrulation indicated by the red fractions (WT: n = 39; dzy: n = 30; rap1: n = 31).

The entire process of VFF is remarkably fast. In order to obtain a quantitative picture of the process of apical constriction and of the defects in dzy and rap1 GLC, we tracked mid-ventral cells in time-lapse recordings of gastrulating embryos. In wild-type, apical cell area is quickly reduced. Within ∼10 minutes all mid-ventral cells are strongly constricted and disappear into the furrow (Fig. 2M,P; supplementary material Movies 1, 2). Concomitant with total apical area reduction, mid-ventral cells become eccentric, i.e. they constrict noticeably less in anterior-posterior than latero-ventral direction (Fig. 2M; supplementary material Movie 2), probably due to asymmetric tension along both axes (Martin et al., 2010).

In dzy GLC, apical constriction is noticeably slowed down in mid-ventral cells, but still causes ventral cells to eventually develop a ventral furrow by reducing apical area and gaining eccentricity similar to wild-type (Fig. 2N,Q; supplementary material Movies 1, 2). Strikingly, although a ventral furrow has finally formed after about double the time compared to wild-type, the furrow fails to close completely (Fig. 1H′; Fig. 2H) and even slightly opens up again (supplementary material Movie 2).

rap1 GLC show a more drastic phenotype. Apical constrictions do not take place evenly in the band of mid-ventral cells, but occur in an apparently disperse pattern. Most cells undergo strong area reduction, while others only constrict weakly, remain entirely unconstricted or become even bloated (Fig. 2I–L,O,R; supplementary material Fig. S3; supplementary material Movies 1, 2). Often even higher constriction levels are reached in comparison to wild-type (Fig. 2P,R, dark blue fraction). However, wild-type cells might reach the same level of constriction, but at a time they have already disappeared into the furrow. Mid-ventral cells in rap1 GLC stretch in apicobasal direction throughout stages 6 and 7 (Fig. 2J–L; see also Fig. 4, panel G6l), but do not gain a markedly eccentric shape (Fig. 2O; supplementary material Movie 1). Despite the strong constriction of most mid-ventral cells in rap1, the ventral epithelium does not form a furrow and the prospective mesoderm is not internalized (Fig. 2I–L; supplementary material Movies 1, 2).

Fig. 4.

Localization of DE-Cad to apical membranes is slowed down in dzy germline clones and largely defective in rap1 germline clones. DE-Cad:GFP in fixed specimens (cross sections in A, A′, D, D′, G, G′; ventral down) and in live embryos (ventral surface views in C, F, I at 5 µm depth) as well as plots of DE-Cad:GFP intensity along membranes going from basal to apical in cross-sections of individual representative cells (B, E, H). (A,A′) In wild-type embryos, DE-Cad is concentrated at the cellularization front during stage 5 (blue arrowheads). At the onset of gastrulation (stage 6e) an accumulation of DE-Cad is visible at the apical side (yellow arrowheads), while the basal accumulation is still present (blue arrowheads). In stage 6l, the basal accumulation has vanished (blue empty arrowheads), while the apical accumulation has developed into a prominent belt around the entire epithelium (yellow arrowheads). (B) The basal peak in stage 5 indicates the localization of DE-Cad to the cellularization front (blue curve). At stage 6e, the basal peak remains and another apical peak comes up (red curve). At stage 6l, the basal peak has disappeared, but intensity strongly peaks in the apical domain (yellow curve). (D,D′) In dzy GLC, DE-Cad localizes normally to the cellularization front during stage 5 (blue arrowheads), but the emergence of an apical accumulation is not evident by stage 6e (yellow empty arrowheads) and is first visible in stage 6l (yellows arrowheads). (E) DE-Cad intensity plots reveal normal basal localization during stage 5, but a missing apical peak in stage 6e (red) and a pertinent basal peak in stage 6l (yellow). (G,G′) In rap1 GLC, DE-Cad localizes normally to the cellularization front during stage 5 (blue arrowheads), but appears diffuse at stage 6e. Apical accumulation only occurs locally at late gastrulation (stage 6l, compare filled with empty arrowheads). Instead, particles of DE-Cad can be detected (red arrowheads). (H) Intensity plots show an unaltered basal concentration during stage 5 (blue), but a highly disperse distribution at stage 6e (red). In stage 6l, many cells still exhibit a disperse DE-Cad distribution, in particular the lack of apical accumulation (yellow). (C,F,I) Stills of time-lapse movies starting at completion of cellularization, i.e. onset of gastrulation. In rap1 GLC, conspicuous large particles of DE-Cad float through the cytoplasm (I, red arrowheads), which are not seen in wild-type embryos (C) and dzy GLC (F). Varying cytoplasmic signal among different cells reflects varying apicobasal nuclear positions as nuclei move basally during the course of gastrulation (Kam et al., 1991); also see panel G6l and supplementary material Fig. S5.

Fig. 4.

Localization of DE-Cad to apical membranes is slowed down in dzy germline clones and largely defective in rap1 germline clones. DE-Cad:GFP in fixed specimens (cross sections in A, A′, D, D′, G, G′; ventral down) and in live embryos (ventral surface views in C, F, I at 5 µm depth) as well as plots of DE-Cad:GFP intensity along membranes going from basal to apical in cross-sections of individual representative cells (B, E, H). (A,A′) In wild-type embryos, DE-Cad is concentrated at the cellularization front during stage 5 (blue arrowheads). At the onset of gastrulation (stage 6e) an accumulation of DE-Cad is visible at the apical side (yellow arrowheads), while the basal accumulation is still present (blue arrowheads). In stage 6l, the basal accumulation has vanished (blue empty arrowheads), while the apical accumulation has developed into a prominent belt around the entire epithelium (yellow arrowheads). (B) The basal peak in stage 5 indicates the localization of DE-Cad to the cellularization front (blue curve). At stage 6e, the basal peak remains and another apical peak comes up (red curve). At stage 6l, the basal peak has disappeared, but intensity strongly peaks in the apical domain (yellow curve). (D,D′) In dzy GLC, DE-Cad localizes normally to the cellularization front during stage 5 (blue arrowheads), but the emergence of an apical accumulation is not evident by stage 6e (yellow empty arrowheads) and is first visible in stage 6l (yellows arrowheads). (E) DE-Cad intensity plots reveal normal basal localization during stage 5, but a missing apical peak in stage 6e (red) and a pertinent basal peak in stage 6l (yellow). (G,G′) In rap1 GLC, DE-Cad localizes normally to the cellularization front during stage 5 (blue arrowheads), but appears diffuse at stage 6e. Apical accumulation only occurs locally at late gastrulation (stage 6l, compare filled with empty arrowheads). Instead, particles of DE-Cad can be detected (red arrowheads). (H) Intensity plots show an unaltered basal concentration during stage 5 (blue), but a highly disperse distribution at stage 6e (red). In stage 6l, many cells still exhibit a disperse DE-Cad distribution, in particular the lack of apical accumulation (yellow). (C,F,I) Stills of time-lapse movies starting at completion of cellularization, i.e. onset of gastrulation. In rap1 GLC, conspicuous large particles of DE-Cad float through the cytoplasm (I, red arrowheads), which are not seen in wild-type embryos (C) and dzy GLC (F). Varying cytoplasmic signal among different cells reflects varying apicobasal nuclear positions as nuclei move basally during the course of gastrulation (Kam et al., 1991); also see panel G6l and supplementary material Fig. S5.

Thus, loss of maternal Dzy results in lagging apical constrictions leading to a considerable slowdown and instability of the ventral furrow, while loss of maternal Rap1 renders apical constriction highly variable among mid-ventral cells and prevents formation of the ventral furrow.

Apical assembly and contraction of the actomyosin filament system is not affected in dzy or rap1 germline clones

For cell shape change to occur, an actomyosin filament system has to assemble and to contract while being tethered to the cell membrane in order to transmit the exerted force. As shown above, apical constrictions during VFF are compromised in dzy and rap1 GLC, so we asked if the actomyosin filament system is properly localized and contractile in ventral cells.

In wild-type, non-muscle myosin heavy chain (MyoII) localizes to the furrow canals during cellularization (Young et al., 1991), compare Fig. 3A. While remaining basally throughout the dorsal and lateral epithelium, MyoII specifically relocalizes to the apex in ventral cells at the onset of VFF. MyoII coalesces, concomitant with apical constriction of these cells, which gives MyoII a dense, contiguous appearance (Fig. 3B–D). After mesodermal cells have been internalized, signal intensity of MyoII quickly vanishes in these cells indicating that the actomyosin is disassembling (Fig. 3E,F).

Fig. 3.

The actomyosin meshwork is contractile in dzy and rap1 germline clones. Localization of MyoII (Zip) in wild-type (A–F), dzy GLC (G–L) and rap1 GLC (M–R) shown in both cross-section and surface section. (A) In wild-type embryos, MyoII is localized to the cellularization front during stage 5. (B) After completion of cellularization, MyoII disappears from the basal side (empty arrowheads) and relocalizes to the apex (arrowhead) in mid-ventral cells. (C) Apical constriction takes place making MyoII appear as a contiguous band along apices of mid-ventral cells (arrowhead). (D) Surface sections show coalesced MyoII within constricted cells. (E,F) After invagination MyoII signal is diminished in internalized cells. (G,M) In dzy and rap1 GLC, localization of MyoII to the cellularization front is unchanged. (H,N) Basal disappearance and apical relocalization in ventral cells also occur normally. (I,J,O,P) MyoII coalesces (arrowheads), but is not in touch with cell membranes. (K,L) In dzy GLC, cells are eventually constricted during late gastrulation, and MyoII appears contiguous. (Q,R) In rap1 GLC, coalesced MyoII appears discontiguous as cell constriction remains incomplete (arrowheads). Depth of surface sections: 2 µm. In dzy and rap1 cross-sections (G, I, K, M, O, Q) the Nrt channel has been omitted for better visualization of MyoII.

Fig. 3.

The actomyosin meshwork is contractile in dzy and rap1 germline clones. Localization of MyoII (Zip) in wild-type (A–F), dzy GLC (G–L) and rap1 GLC (M–R) shown in both cross-section and surface section. (A) In wild-type embryos, MyoII is localized to the cellularization front during stage 5. (B) After completion of cellularization, MyoII disappears from the basal side (empty arrowheads) and relocalizes to the apex (arrowhead) in mid-ventral cells. (C) Apical constriction takes place making MyoII appear as a contiguous band along apices of mid-ventral cells (arrowhead). (D) Surface sections show coalesced MyoII within constricted cells. (E,F) After invagination MyoII signal is diminished in internalized cells. (G,M) In dzy and rap1 GLC, localization of MyoII to the cellularization front is unchanged. (H,N) Basal disappearance and apical relocalization in ventral cells also occur normally. (I,J,O,P) MyoII coalesces (arrowheads), but is not in touch with cell membranes. (K,L) In dzy GLC, cells are eventually constricted during late gastrulation, and MyoII appears contiguous. (Q,R) In rap1 GLC, coalesced MyoII appears discontiguous as cell constriction remains incomplete (arrowheads). Depth of surface sections: 2 µm. In dzy and rap1 cross-sections (G, I, K, M, O, Q) the Nrt channel has been omitted for better visualization of MyoII.

In dzy and rap1 GLC, the specific ventral relocalization of MyoII is unchanged: Like in wild-type, MyoII is present at the furrow canals during cellularization (Fig. 3G,M), before it specifically disappears from the basal side in ventral cells and localizes to the apex (Fig. 3H,I,N,O). The localization of the other structural component of the actomyosin filament system, F-Actin, is also unchanged in dzy and rap1 GLC (data not shown). Furthermore, RhoGEF2, the key component conferring contractile actomyosin, is properly localized to the apices of ventral cells at early gastrulation in dzy and rap1 GLC (supplementary material Fig. S4). In rap1 GLC, MyoII coalesces within mid-ventral cells during stage 6m (Fig. 3O,P, arrowheads). However, unlike wild-type (Fig. 3C,D), this coalesced MyoII appears largely detached from cell membranes since many cells are unconstricted. Also in stage 6l coalesced MyoII is visible in unconstricted or incompletely constricted cells (Fig. 3Q,R, arrowheads), resulting in a discontiguous and torn impression. This coalesced MyoII indicates that the actomyosin filaments have contracted. Apparently not in all cells, however, does the actomyosin contraction efficiently pull along the membranes indicating compromised attachment of the cytoskeleton to cell membranes. Similar conclusions have been drawn in other studies which revealed coalesced MyoII in unconstricted cells (Dawes-Hoang et al., 2005; Martin et al., 2010; Sawyer et al., 2009). Also in dzy GLC, coalesced MyoII is evident within unconstricted cells at stage 6m (Fig. 3I,J, arrowheads). Only at later gastrulation cells finally constrict, and MyoII shows a contiguous appearance (Fig. 3K,L) that is similar to wild-type at an earlier time point of VFF.

Thus, we conclude that the actomyosin filament system is contractile in both dzy and rap1 GLC since MyoII does coalesce. In dzy GLC, however, actomyosin possibly fails to properly attach to cell membranes before late gastrulation. In rap1 GLC, attachment seems to vary among mid-ventral cells, ranging from sufficiently functional to defective, resulting in a broad distribution of achieved constriction levels (Fig. 2R).

Accumulation of DE-Cad into an apical belt depends on Dzy and Rap1

Adherens junctions (AJs) in epithelia are aggregates of DE-Cadherin (DE-Cad) and other proteins that are coupled to the actin cytoskeleton, provide a membrane-anchor to the contractile actomyosin and allow cells to undergo shape changes. As actomyosin is correctly localized and appears functional, but does not seem capable of properly constricting ventral cells in dzy and rap1 GLC, we investigated junction formation by means of DE-Cad:GFP, in both living and fixed specimens. During gastrulation DE-Cad undergoes characteristic changes in localization which occur likewise in ventral, lateral and dorsal cells. Due to the lack of morphological change, documentation is facilitated in lateral and dorsal cells (Fig. 4). The analogous events in ventral cells are documented in supplementary material Fig. S6. In addition to the descriptions of qualitative changes in localization, we also measured pixel intensities along membranes going from basal to apical in order to put our observations about DE-Cad localization onto a quantitative basis (Fig. 4B,E,H; supplementary material Fig. S6C,F,I).

Like MyoII, DE-Cad localizes to the ingressing cellularization front during stage 5 in wild-type, thus appearing basal in late stage 5 cross-sections (Fig. 4, panels A5, A′5, B). Towards the end of cellularization DE-Cad starts to prominently accumulate in the apical domain around the circumference of the epithelium indicating the establishment of the apical adhesion belt at the onset of gastrulation (Fig. 4, panels A6e, A′6e, B; supplementary material Fig. S6A,C). As gastrulation proceeds, the remaining basal DE-Cad accumulation dissolves while the apical belt grows more pronounced (Fig. 4, panels A6l, A′6l, B). Concomitantly with this apical junction assembly mid-ventral cells constrict their apices indenting the ventral epithelium (Fig. 4, panel A6e; supplementary material Fig. S6B,C). In previous studies Arm has also been shown to accumulate apically at the end of cellularization consistent with the set-up of the apical junction belt immediately prior to gastrulation (Dawes-Hoang et al., 2005; Hunter and Wieschaus, 2000). Ventral cells exhibit the same apical localization of DE-Cad at the end of cellularization as do lateral and dorsal cells (supplementary material Fig. S6A–C); however, possibly since the apical cortex commences flattening in these cells, the dome-like eversions are missing in ventral cells letting DE-Cad appear more apical than in lateral and dorsal cells (compare ventral to lateral and dorsal in Fig. 4, panels A6e, A′6e to supplementary material Fig. S6A; Fig. 4, panel A′6l to supplementary material Fig. S6B). Also, the basal accumulation is lost earlier in ventral cells (compare Fig. 4, panel A′6e to supplementary material Fig. S6A).

In dzy GLC, DE-Cad localizes basally during stage 5 in accordance with cellularization proceeding normally (Fig. 4, panels D5, D′5, E; supplementary material Fig. S7). However, at the onset of gastrulation the establishment of apical junctions is delayed in the entire epithelium as apical accumulation of DE-Cad has not taken place by stage 6e (Fig. 4, panels D6e, D′6e, E; supplementary material Fig. S6D,F; supplementary material Fig. S7) and is visible no earlier than stage 6l, accompanied by the first apical constrictions and tissue indentation (Fig. 4, panels D6l, D′6l, E; supplementary material Fig. S6E,F; supplementary material Fig. S7).

In rap1 GLC, DE-Cad also localizes basally to the cellularization front (Fig. 4, panels G5, G5′, H), but in addition, striking punctate ectopic accumulations of DE-Cad are evident in the cells (Fig. 4G,G′; supplementary material Fig. S6G,H, red arrowheads). Also in confocal live-imaging peculiar DE-Cad:GFP rich particles can be seen floating through the cytoplasm in ventral, lateral and dorsal cells of the epithelium that are not found in wild-type or dzy GLC (Fig. 4C,F,I; supplementary material Movie 4). DE-Cad appears largely dispersed along lateral membranes and only fragmentary apical accumulation is evident, but a complete circumferential apical belt is not established (Fig. 4G,G′,H; supplementary material Fig. S6G–I). Interestingly, accumulation of the minor AJ component Arm (β-catenin) into an apical belt appears to be unaffected (supplementary material Fig. S8).

We conclude that Dzy and Rap1 are not required for localizing DE-Cad to AJs at the cellularization front, but for its accumulation into an apical belt at the onset of gastrulation. This process is considerably slowed down in dzy GLC and largely defective in rap1 GLC. As a consequence, the cytoskeleton cannot attach to cell membranes during early gastrulation in dzy GLC since AJs are not in place yet. Thus, first actomyosin contractions are without effect on cell constriction. In rap1 GLC, the AJ belt remains incomplete resulting in a varying capability of attaching the cytoskeleton, so actomyosin contractions lead to a varying degree of constriction among mid-ventral cells (Fig. 2R; Fig. 3).

Dzy and Rap1 both localize cortically prior to ventral furrow formation, but are strongly reduced in invaginated cells

Dzy and Rap1 have both been found to localize to AJs during later developmental events (Boettner and Van Aelst, 2007; Knox and Brown, 2002; Wang et al., 2006). Here, we were interested in the subcellular localization of Dzy and Rap1 immediately prior to and during gastrulation. As gastrulation starts, both proteins are distributed along the lateral cell membranes in all cells of the epithelium but concentrate in the cortex (Fig. 5A,A′,F,F′) in agreement with a role in apical AJ establishment as well as previous findings (Sawyer et al., 2009). Intriguingly, while this cortical enrichment remains present in the ectoderm throughout and after gastrulation, both Dzy:GFP and Rap1:GFP signals start to fade in mesodermal cells as soon as these have been internalized in stage 7 (Fig. 5B,B′,G,G′,K). Later the signal intensity further decreases in mesodermal cells, but is maintained in the ectoderm (Fig. 5C–E′,H–J′). By stage 10 the relative mesodermal intensity has fallen by 40% for Dzy and by 80% for Rap1 (Fig. 5K), indicating that both protein levels decrease considerably in the mesoderm after VFF has been accomplished.

Fig. 5.

Both Dzy and Rap1 localize to the cortex during ventral furrow formation, but fade in internalized cells. (A–J′) Localization of Dzy:GFP (A–E′) and Rap1:GFP (F–J′) during stages 6–10, detected by anti-GFP staining. At the onset of VFF, both Dzy and Rap1 are concentrated in the cortex of cells around the entire epithelium (A,A′,F,F′). As soon as the mesoderm has been invaginated in stage 7, the level of both proteins starts to fade in cells that have completed internalization (B,B′,G,G′, empty vs. filled arrowheads). Intensity further declines during stages 8–10 when the mesodermal tube collapses and cells migrate dorsally on the ectoderm (C–J′). Rap1 shows a stronger decrease compared with Dzy, rendering the mesoderm almost invisible in the GFP channel (arrowheads in J,J′). Both proteins remain strongly concentrated in the cortices of ectodermal cells. (K) Graph showing the mean GFP intensities of Dzy and Rap1 in the mesoderm normalized to the ectoderm. These are essentially at an 1:1 ratio at stage 6, but have strongly declined by stage 10.

Fig. 5.

Both Dzy and Rap1 localize to the cortex during ventral furrow formation, but fade in internalized cells. (A–J′) Localization of Dzy:GFP (A–E′) and Rap1:GFP (F–J′) during stages 6–10, detected by anti-GFP staining. At the onset of VFF, both Dzy and Rap1 are concentrated in the cortex of cells around the entire epithelium (A,A′,F,F′). As soon as the mesoderm has been invaginated in stage 7, the level of both proteins starts to fade in cells that have completed internalization (B,B′,G,G′, empty vs. filled arrowheads). Intensity further declines during stages 8–10 when the mesodermal tube collapses and cells migrate dorsally on the ectoderm (C–J′). Rap1 shows a stronger decrease compared with Dzy, rendering the mesoderm almost invisible in the GFP channel (arrowheads in J,J′). Both proteins remain strongly concentrated in the cortices of ectodermal cells. (K) Graph showing the mean GFP intensities of Dzy and Rap1 in the mesoderm normalized to the ectoderm. These are essentially at an 1:1 ratio at stage 6, but have strongly declined by stage 10.

Thus, Dzy and Rap1 are similarly localized cortically in the epithelium from stage 6 onwards consistent with being involved in the set-up and possibly in the maintenance of epidermal apical AJs. Conversely, both proteins are substantially reduced in internalized cells of the prospective mesoderm. This raises the question whether the down-regulation of Dzy and Rap1 is relevant for further development of the mesoderm, in particular for its mesenchymal properties.

The diminishment of Dzy and Rap1 is required to allow mesodermal migration

After completion of VFF the internalized mesoderm quickly loses its epithelial character and undergoes epithelial-to-mesenchymal transition (EMT): The invaginated tube collapses and cells lose their tight adhesion to each other. After EMT, the cells migrate dorsally on the ectoderm, forming a monolayer during stage 9 to 10 (Fig. 5C–E; Fig. 6A). This process is accompanied by the loss of AJs and the diminishment of DE-Cad (Oda et al., 1998).

Fig. 6.

Overexpression of Dzy or constitutively active Rap1 impairs mesodermal migration. (A–G) Morphology of the invaginated mesoderm (marked by Twi) in stage 10 of wild-type embryos, dzy GLC, rap1 GLC and upon overexpression of Dzy or constitutively active Rap1V12 by means of da::gal4 or twi::gal4 drivers (for better visual assessment of DE-Cad intensity supplementary material Fig. S9 shows panels A–G without the red channel). (A) In wild-type, the invaginated tube has collapsed and mesodermal cells have spread dorsally along the ectoderm forming a monolayer. (B) In dzy GLC, the mesoderm has spread dorsally and formed a normal monolayer, although a minor part has not been internalized (yellow arrowhead). (C) In rap1 GLC, the majority of the mesoderm is not internalized. A small internalized fraction shows unaffected monolayer formation (blue arrowhead). (D,E) When Dzy is overexpressed ubiquitously (D) or in the prospective mesoderm only (E), the mesoderm fails to fully spread and often does not form a monolayer. (F,G) Upon overexpression of Rap1V12 this phenotype is even stronger; the mesoderm shows pronounced clumping and defective monolayer formation. (H) Graph showing the DE-Cad intensity in the mesoderm normalized to that in the ectoderm for the genotypes shown in A–G. Rap1V12 overexpression yields significantly elevated levels of DE-Cad intensity in mesodermal cells. (I,J) Quantification of the non-monolayer phenotypes upon Dzy or Rap1V12 overexpression. Mesodermal spreading was measured by relating the mesodermal extension along the ectoderm (a) to the full stretch from the ventral-most point to the amnioserosa within the ectoderm (b). Mesodermal clumping was quantified by relating the thickness of the mesoderm (c) to the thickness of mesoderm plus ectoderm (d) above the ventral-most point of the embryo.

Fig. 6.

Overexpression of Dzy or constitutively active Rap1 impairs mesodermal migration. (A–G) Morphology of the invaginated mesoderm (marked by Twi) in stage 10 of wild-type embryos, dzy GLC, rap1 GLC and upon overexpression of Dzy or constitutively active Rap1V12 by means of da::gal4 or twi::gal4 drivers (for better visual assessment of DE-Cad intensity supplementary material Fig. S9 shows panels A–G without the red channel). (A) In wild-type, the invaginated tube has collapsed and mesodermal cells have spread dorsally along the ectoderm forming a monolayer. (B) In dzy GLC, the mesoderm has spread dorsally and formed a normal monolayer, although a minor part has not been internalized (yellow arrowhead). (C) In rap1 GLC, the majority of the mesoderm is not internalized. A small internalized fraction shows unaffected monolayer formation (blue arrowhead). (D,E) When Dzy is overexpressed ubiquitously (D) or in the prospective mesoderm only (E), the mesoderm fails to fully spread and often does not form a monolayer. (F,G) Upon overexpression of Rap1V12 this phenotype is even stronger; the mesoderm shows pronounced clumping and defective monolayer formation. (H) Graph showing the DE-Cad intensity in the mesoderm normalized to that in the ectoderm for the genotypes shown in A–G. Rap1V12 overexpression yields significantly elevated levels of DE-Cad intensity in mesodermal cells. (I,J) Quantification of the non-monolayer phenotypes upon Dzy or Rap1V12 overexpression. Mesodermal spreading was measured by relating the mesodermal extension along the ectoderm (a) to the full stretch from the ventral-most point to the amnioserosa within the ectoderm (b). Mesodermal clumping was quantified by relating the thickness of the mesoderm (c) to the thickness of mesoderm plus ectoderm (d) above the ventral-most point of the embryo.

Dzy and Rap1 are involved in the establishment and possibly the maintenance of the apical adhesion belt in epithelial cells, but on the other hand their expression level quickly decreases in the more motile mesodermal cells. Therefore, we wondered if the diminishment of Dzy or Rap1 is a requirement for EMT and subsequent mesenchymal migration. We ectopically expressed Dzy or the constitutively active form Rap1V12 ubiquitously or specifically in the mesoderm. In both cases no effects were observed prior to or during VFF, but after EMT subsequent migration and monolayer formation of mesenchymal cells are affected considerably. Mesodermal cells show clumping and reduced spreading by stage 10, causing the cells to line up in several layers (Fig. 6D–G,I,J). In case of ectopic expression of constitutively active Rap1 these morphogenetic defects are accompanied by significantly elevated relative levels of DE-Cad in mesenchymal cells (Fig. 6H). Thus, ectopic activity of Rap1 maintains DE-Cad and thereby possibly strengthens cell–cell adhesion. We did not find mesodermal migration to be defective in dzy GLC (Fig. 6B) or, unlike previously reported (McMahon et al., 2010), in rap1 GLC (Fig. 6C). As observed in other VFF mutants, the mesoderm has the capability to move inside the embryo even though VFF is affected (Seher et al., 2007; Seher and Leptin, 2000). This indicates that a normal invagination is not essential for mesoderm internalization. Indeed, in dzy GLC the bulk of the mesoderm has been internalized by stage 10, in spite of the VFF defects, and shows normal monolayer formation. Only a minor fraction remains outside (Fig. 6B, arrowhead). In rap1 GLC the ventral furrow phenotype is more severe, the overwhelming majority of the prospective mesoderm has not been internalized and broadly separates right and left of the neuroectodermal primordia (Fig. 6C). Only a small fraction has reached the interior, but this fraction seems to have normal mesenchymal properties as cells form a regular monolayer (Fig. 6C, arrowhead).

We propose that ectopic presence of Dzy or of constitutively active Rap1 in invaginated mesodermal cells is sufficient to affect mesenchymal development by impairing cell spreading, possibly due to increased cell–cell adhesion. We deduce that the reduction of Dzy and active Rap1 is necessary for mesodermal cells to allow for efficient migration and monolayer formation.

Discussion

We found that both the PDZ-GEF Dzy and its target Rap1 are required for ventral furrow formation (VFF) during Drosophila gastrulation. In the absence of Dzy the establishment of the circumferential adhesion belt is slowed down while in the absence of Rap1 only a fragmentary adhesion belt is formed (Fig. 4). In the case of dzy, this slowdown in apical junction assembly translates into a slowdown of apical cell constriction, since the cytoskeleton cannot attach to membranes during early gastrulation. Thus, first actomyosin contractions do not evoke cell shape change (Figs 2, 3, 7). In the case of rap1, junction assembly is much more severely affected, and only a fragmentary apical junction belt forms during gastrulation (Fig. 4). This results in a variable capability of mid-ventral cells to undergo constriction since only variable levels of apical AJs are available to connect to the contracting actomyosin (Figs 2, 3, 7). Dzy and Rap1 localize cortically during and after gastrulation consistent with a role in junction assembly and possibly maintenance. Levels of both proteins are diminished in the mesoderm once it has been internalized (Fig. 5). Overexpressing Dzy or Rap1V12 in the mesoderm results in an inhibition of mesodermal spreading (Fig. 6), possibly by keeping up DE-Cad-mediated adhesion. Our findings underline the roles of the PDZ-GEF Dzy and its GTPase Rap1 as critical factors regulating the dynamics of adherens junction (AJ) formation in Drosophila gastrulation.

Fig. 7.

Model of cell shape changes in wild-type, dzy germline clones and rap1 germline clones during ventral furrow formation. Schematic depiction of mid-ventral cells in apical surface views (L: lateral; M: medial; A: anterior; P: posterior). Wild-type: late stage 5, towards completion of cellularization, DE-Cad begins to accumulate in apical AJs, and the actomyosin filament system starts to assemble. 6e, the contracting actomyosin has attached to a belt of mature AJs and started to constrict apical cell area. 6m/6l, actomyosin contractions continue and cell apices are further constricted. dzy GLC: late 5, apical actomyosin assembly begins, but the establishment of apical AJs is delayed. 6e, the actomyosin has readily assembled but does not find sites to attach to. Thus, initial contractions do not induce cell shape change. 6m, apical AJs begin to form providing weak attachment sites. Thus, first cell constrictions can be evoked by the contracting actomyosin. 6l, the apical AJ belt has finally formed, so actomyosin contraction can be fully translated into cell constriction, eventually. rap1 GLC: late 5, apical actomyosin begins to assemble, but DE-Cad is not accumulated apically. In addition, stable integration of DE-Cad in cell membranes is affected so particles of DE-Cad float through the cytoplasm. 6e, actomyosin assembly is completed and contractions start, but are not translated into cell shape change as apical AJs are missing. 6m, apical AJs begin to form in some cells generating attachment sites for the contracting actomyosin. Thus, apical constriction can commence in a subset of ventral cells, while others remain unconstricted. 6l, ventral cells have constricted according to the number of available apical AJs leading to a random arrangement of constricted and unconstricted cells. Constricted cells have achieved a high level of constriction since neighbouring unconstricted cells have not exerted a counterforce.

Fig. 7.

Model of cell shape changes in wild-type, dzy germline clones and rap1 germline clones during ventral furrow formation. Schematic depiction of mid-ventral cells in apical surface views (L: lateral; M: medial; A: anterior; P: posterior). Wild-type: late stage 5, towards completion of cellularization, DE-Cad begins to accumulate in apical AJs, and the actomyosin filament system starts to assemble. 6e, the contracting actomyosin has attached to a belt of mature AJs and started to constrict apical cell area. 6m/6l, actomyosin contractions continue and cell apices are further constricted. dzy GLC: late 5, apical actomyosin assembly begins, but the establishment of apical AJs is delayed. 6e, the actomyosin has readily assembled but does not find sites to attach to. Thus, initial contractions do not induce cell shape change. 6m, apical AJs begin to form providing weak attachment sites. Thus, first cell constrictions can be evoked by the contracting actomyosin. 6l, the apical AJ belt has finally formed, so actomyosin contraction can be fully translated into cell constriction, eventually. rap1 GLC: late 5, apical actomyosin begins to assemble, but DE-Cad is not accumulated apically. In addition, stable integration of DE-Cad in cell membranes is affected so particles of DE-Cad float through the cytoplasm. 6e, actomyosin assembly is completed and contractions start, but are not translated into cell shape change as apical AJs are missing. 6m, apical AJs begin to form in some cells generating attachment sites for the contracting actomyosin. Thus, apical constriction can commence in a subset of ventral cells, while others remain unconstricted. 6l, ventral cells have constricted according to the number of available apical AJs leading to a random arrangement of constricted and unconstricted cells. Constricted cells have achieved a high level of constriction since neighbouring unconstricted cells have not exerted a counterforce.

Dzy guarantees the fast assembly of the apical junction belt required for ventral furrow formation

Apical constriction of ventral cells is known to be a major driving force of VFF and much progress has been made in deciphering the signal cascade leading from ventral fate determinants to an assembly of a contractile actomyosin at the apices of ventral cells (Barrett et al., 1997; Dawes-Hoang et al., 2005; Häcker and Perrimon, 1998; Kölsch et al., 2007; Morize et al., 1998; Parks and Wieschaus, 1991). Also, the importance of tight coupling between the contracting actomyosin and the cell membranes mediated by AJs has been previously highlighted (Sawyer et al., 2009). Although MyoII has been implicated as a downstream target of dzy during dorsal closure (Boettner and Van Aelst, 2007), we cannot attribute the slowdown in cell shape change during VFF seen in dzy GLC to a slowdown in apical assembly of the actomyosin apparatus. Unlike what has been reported for dorsal closure, actomyosin exhibits the same relocalization to the apex of ventral cells at the end of cellularization in dzy GLC and in wild-type. Furthermore, we find MyoII coalesced into balls within unconstricted cells when gastrulation starts, supporting the notion of a contracting actomyosin meshwork (Fig. 3). Coalesced MyoII within unconstricted cells has also been reported previously for ventral cells in arm (Dawes-Hoang et al., 2005), cno and rap1 GLC (Sawyer et al., 2009) all of which exhibit defective cell constriction. In these studies, this observation was considered an indication of contracting actomyosin that is detached from cell membranes. Our findings are consistent with this view (Figs 3, 4).

Previous work has revealed that ventral cells are not constricted by continuous contraction and that circumferential actomyosin cables do not contribute significantly to the constriction. Instead, a medially localized actomyosin meshwork is thought to reach out to make contact to AJs at the cell membranes and executes discontinuous contraction pulses to constrict the apex (Martin et al., 2009). Our observation that apical constriction still occurs in dzy GLC, later than in wild-type, but apparently as soon as AJs are in place, are in accordance with these findings. Thus, apical constriction is not irrecoverably affected if AJs are not ready at the onset of gastrulation. Actomyosin contraction appears to take place in a dynamical and repeated pulsed fashion over the entire time-span of gastrulation allowing cells to constrict eventually, despite an initial delay in AJ formation (Fig. 3J,L; Fig. 7).

A puzzling feature of the dzy phenotype is the failure of the ventral furrow to finally close although ventral cells have undergone complete, albeit delayed, apical constriction (Fig. 2M,N; supplementary material Movie 3). We propose that the invagination of the mesoderm has to occur within a critical time slot, which is missed in dzy GLC due to the delay in AJ establishment and, consequently, apical cell constriction. In fact, the ventral furrow of dzy GLC very much resembles the ventral furrow of a wild-type embryo 5 to 10 minutes earlier (compare frame 51 to frame 88 in supplementary material Movie 2; 10 to 15 in supplementary material Movie 3). Still, the furrow is not properly sealed in the end, less tissue moves inside and often the furrow opens up again (supplementary material Movie 3). This supports the notion that apical constriction alone is not sufficient to internalize the ventral furrow. Computer simulations have indicated that apical constriction alone is incapable of generating a tissue invagination and have postulated ectodermal pushing as a second source of force to internalize the ventral furrow (Conte et al., 2009). Such a force could be exerted by turgor pressure in medio-lateral direction within the cellular blastoderm. The ventral furrow may serve as a ‘predetermined breaking line’, where the tissue can give in to the inherent pressure. The delay in VFF of dzy GLC leads to a temporal overlap with germband extension and PMG invagination that immediately follow the internalization of the ventral furrow in wild-type (Campos-Ortega and Hartenstein, 1985). Both processes are likely to reduce the epithelial pressure in medio-lateral dimension since they expand the epithelium in the antero-posterior dimension. Consequently, pressure might have already become too low to generate the force required to push in the mesoderm when the ‘breaking line’ has finally emerged. In addition, it cannot be ruled out that the ventral furrow is not properly closed and opens up again in dzy GLC because of a failure in sealing the edges of the furrow.

Rap1 ensures membrane association and apical accumulation of DE-Cad

In contrast to dzy GLC, only a fragmentary AJ belt is formed in rap1 GLC as DE-Cad is diffusely distributed in the membranes and shows delayed and incomplete apical accumulation (Fig. 4). In addition, DE-Cad reveals a striking cytoplasmic mislocalization to floating particles that are seen in rap1 GLC only (Fig. 4; supplementary material Movie 4). Although the nature of these particles remains to be clarified, we speculate they represent DE-Cad rich membrane vesicles originating from the cell membrane. It has been reported earlier (Sawyer et al., 2009) that initial AJ assembly is unaffected in rap1 GLC, but this conclusion was based on anti-Arm staining which look unaffected in our analysis as well (supplementary material Fig. S8). Thus, Rap1 seems to act on DE-Cad specifically to assure its proper localization. In mammalian cells regulation of DE-Cad endocytosis has long been recognized as a cellular mechanism to modulate AJs (de Beco et al., 2009; Le et al., 1999). In this context, Rap1 has been implicated in having a key role in stabilizing DE-Cad in membrane-bound aggregates as it is thought to enhance binding of DE-Cad to p120-catenin, which may serve as a cap protecting DE-Cad from being endocytosed (Hogan et al., 2004; Hoshino et al., 2005). On the other hand, p120-catenin appears to play only a minor role in Drosophila (Fox et al., 2005; Myster et al., 2003; Pacquelet et al., 2003). Despite the accordance with previous studies (Sawyer et al., 2009), the unaffected apical accumulation of Arm in rap1 GLC (supplementary material Fig. S8) was surprising, especially since loss of DE-Cad is reported to entail loss of Arm in various tissues (Tepass et al., 1996). However, Arm is also involved in many other DE-Cad independent processes, e.g. acting as a signal molecule or transcription factor, so a requirement of DE-Cad for its localization does not appear coercive.

Albeit the precise mechanism remains to be investigated, we assume that in the absence of maternal Rap1, confinement of DE-Cad to cell membranes and accumulation into stable apical junctions is severely compromised. Instead, only fragmentary junctions are formed whose stability may vary stochastically. Thus, AJ fragmentation may affect different cells to a different degree. As a consequence, ventral cells show a broad distribution of constriction capability ranging from complete constriction to a total failure of constriction (Fig. 2O,R; Fig. 7). It may be recognized that apical constriction does not appear to be slowed down in those cells of rap1 GLC that are capable of undergoing constriction (Fig. 2O,R; supplementary material Movie 2). A reason for this could be the lack of constriction in surrounding cells, so constricting cells experience considerably less opposing force from their neighbours in the epithelium. This could allow them to constrict faster and make up the inefficient actomyosin attachment in their membranes. Similarly, the lack of constriction in neighbours may allow constricting cells to constrict uniformly (‘isotropically’), rather than become eccentric like wild-type cells (Fig. 2M,O). Due to the discontiguous actomyosin meshwork in the ventral epithelium, tension in the anteroposterior axis will be strongly reduced so constricting cells are not forced into an eccentric morphology. Indeed, previous work has shown that mid-ventral cells can undergo isotropic constriction when anteroposterior tension is disrupted by inflicting tears upon the supracellular actomyosin meshwork (Martin et al., 2010). Surprisingly, in spite of the large fraction of mid-ventral cells with high constriction levels, rap1 GLC do not form a ventral furrow. We assume that the minor fraction of unconstricted and bloated mid-ventral cells has an inhibitory influence on VFF, possibly by interrupting the ‘predetermined breaking line’.

Thus, rap1 and dzy differ qualitatively in their maternal phenotypes because loss of Dzy only delays establishment of AJs whereas loss of Rap1 additionally entails a fragmentation of the AJ belt and massive cytoplasmic mislocalization of DE-Cad. This discrepancy is not in conflict with the concept of Dzy acting exclusively via Rap1, but strongly argues in favour of Rap1 being regulated by additional GEFs besides Dzy (see below).

A possible general role of Dzy and Rap1 in junction formation and maintenance

It must be emphasized that the effects on AJ assembly seen in dzy and rap1 GLC are not confined to the prospective mesoderm but occur around the entire epithelium consistent with the localization of Dzy and Rap1 in wild-type (Fig. 5). dzy and rap1 have been recognized as ‘ventral furrow mutants’ because apical constriction of ventral cells is the earliest process in embryogenesis requiring a properly built apical AJ belt.

With the apical adhesion belt being a prominent feature of ectodermal cells, internalized mesodermal cells show substantially weaker DE-Cad intensity (Oda et al., 1998; Fig. 6A,H) indicating that junctions are disassembled in order to reduce cell-cell adhesion and allow mesenchymal migration. Overexpression of Dzy or Rap1V12 impairs this mesenchymal migration significantly, Rap1V12 noticeably stronger than Dzy alone (Fig. 6H,I,J). This is very plausible given that Dzy works via Rap1 which is considerably reduced in the internalized mesoderm (Fig. 5K). Migration defects upon Rap1V12 overexpression are accompanied by significantly risen relative amounts of DE-Cad in mesenchymal cells (Fig. 6) suggesting the possibility that the downregulation of Rap1 is required to allow AJs to become disassembled in the mesoderm. Accordant results have been found in the Drosophila testis where reduction of AJs can be restored to wild-type level through overexpression of constitutively active Rap1 (Wang et al., 2006). It remains to be seen by what mechanism AJs are disassembled in the internalized mesoderm and how the remarkably fast diminishment of Dzy and Rap1 is triggered. Conceivably, processes accompanying EMT such as mechanical alterations in the cytoskeleton could trigger degradation signals since these processes have been found to have potential signalling ability in other systems (Howard et al., 2011).

How are Dzy and Rap1 embedded in the cascade relaying cell signalling to morphogenesis?

As discussed above, the discrepancy between the maternal phenotypes of dzy and rap1 implies the necessity of other GEFs acting on Rap1 during gastrulation. C3G is a tempting candidate as it has been shown to interact with Rap1 in mammalian cell culture as well as in Drosophila (Dupuy et al., 2005; Ishimaru et al., 1999). Furthermore, it exhibits GEF activity on Drosophila Rap1 in vitro (Shirinian et al., 2010).

In addition to uncovering alternative activators of Rap1 it will be interesting to identify players upstream of Dzy. Despite its cyclic nucleotide binding domain there is no indication so far that Dzy is activated by cAMP signalling (Kuiperij et al., 2003; Pham et al., 2000). However, like several proteins involved in cell polarity, Dzy bears a PDZ domain through which it possibly binds to a membrane scaffold typically involved in mediating quick linkage between signalling molecules and structural proteins (Bilder, 2001). Indeed, the PDZ protein MAGI-1 has been shown to serve as a scaffold for the vertebrate homologue of Dzy (Mino et al., 2000) and is a good candidate for a protein giving the relevant spatial cue. Unravelling the architecture of such a signalling scaffold will be key to understanding how an epithelium can be reorganized so rapidly to allow the extraordinarily fast morphogenesis of the ventral furrow.

Materials and Methods

Fly stocks and generation of germline clones (GLC)

dzyΔ1, dzyΔ8, dzyΔ12 (Huelsmann et al., 2006), dzy-GFP (dPDZ-GEFEGFP) and UAS-dzy (UAS-dPDZ-GEFEGFP) (Boettner and Van Aelst, 2007), Df(2L)ED380 (Ryder et al., 2007), spider-gfp (Morin et al., 2001), DE-Cad:GFP (ubi-DE-Cad-GFP,Oda and Tsukita, 2001), hsFLP1.22; FRT2L-40A and hsFLP1.22; ovoD13L-2X48 FRT3L-2A (Chou et al., 1993; Chou and Perrimon, 1992), da-gal4 (Hinz et al., 1994), twi-gal4 (Giebel et al., 1997), UAS-rap1V12 (Boettner et al., 2003); rap1-gfp and rap1P[5709]FRT3L-2A (Knox and Brown, 2002).

The analysis of dzy GLC was performed with either of the dzy alleles, since no difference was visible in their phenotypes. They are considered dzy nulls since GLC of the deficiency Df(2L)ED380 show the same early embryonic phenotype (not shown). The DE-Cad:GFP fusion protein provides full DE-Cadherin function and does not elicit any adverse effects on embryogenesis or viability (Oda and Tsukita, 2001). rap1P[5709] is an insertion of P{lacW} into rap1 deleting the C-terminal 37 amino acids including the GTP binding sequence and the prenylation site (N. Brown, personal communication). GLC were produced according to standard methods (Chou and Perrimon, 1992).

Staging of embryos

Developmental stages were distinguished following as described by Campos-Ortega and Hartenstein (Campos-Ortega and Hartenstein, 1985). For the fine temporal staging during VFF we distinguished an early (6e), a middle (6m) and a late stage 6 (6l) predominantly based on the position of the pole cells, as documented in supplementary material Fig. S2.

Immunohistochemistry and in situ hybridization

Embryos were fixed by heat-methanol treatment (modified after Müller and Wieschaus, 1996) or by 3.7% formaldehyde according to standard procedures. Primary antibodies: mouse anti-Arm (N2 7A1; 1∶100; DSHB); rat anti-DE-Cad (DCAD2; 1∶100; DSHB); sheep anti-DIG-AP (1∶2000; Roche); rabbit anti-Eve (1∶2000; Frasch et al., 1987); rabbit anti-GFP (1∶2000; Molecular Probes); rabbit anti-Myosin heavy chain (1∶500 of a 5 mg/ml stock; Kiehart and Feghali, 1986); mouse anti-Nrt (BP106; 1∶10; DSHB); rabbit anti-Sna (1∶2000; Reuter and Leptin, 1994); rabbit anti-Twi (1∶5000; Roth et al., 1989); chicken anti-Zip/MyoII (1∶500; Kiehart and Feghali, 1986); rabbit anti-RhoGEF2 (1∶6000; Grosshans et al., 2005). Secondary antibodies, labelled with Alexa 488 (Molecular Probes), Cy3 or biotin (Jackson Labs), were used at 1∶500. DAPI (1∶3000 of a 5 mg/ml stock; Sigma) was used to visualize nuclei. The signal of the biotinylated antibodies was enhanced using Vectastain ‘ABC Elite Kit’ (1∶100; Vector Laboratories) and detected with H2O2 (0.003%; Fluka) and 3,3′-diaminobenzidine (DAB, 0.25 mg/ml; Sigma) for a brown staining, with the same reagents in combination with 0.01% NiCl2 and 0.01% CoCl2 to create a blue staining. Alexa 594-conjugated phalloidin (Molecular Probes) was used at 1 U/ml.

In situ hybridization was performed essentially as described in Tautz and Pfeifle (Tautz and Pfeifle, 1989). DIG-labelled pannier (pnr) (Ramain et al., 1993) and huckebein (hkb) (Brönner et al., 1994) RNA probes were generated from full-length cDNA clones in pNB40 (Brown and Kafatos, 1988).

Fluorescently labelled cross sections were mounted in Aqua-PolyMount (Polysciences). For cross sectioning embryos were cut manually with a syringe needle. Sections were imaged at 50% egg length. Confocal images were taken on a Leica SP2 confocal system on an inverted Leica DM IRBE microscope. For bright-field microscopy embryos were mounted in araldite and analyzed on an Axioplan 2 (Zeiss) equipped with a ProgRes C14 camera (Jenoptik).

Live-imaging

For time-lapse recordings dechorionated embryos were glued with their dorsal side to a slide (fluorescence) or positioned with their lateral side on a slide (bright-field). They were mounted with a drop of water-saturated 3S Voltalef oil under a cover slip supported by two cover slips on each side. Image series were recorded using a Leica SP2 confocal system for fluorescence or a Zeiss Axioplan 2 equipped with a SPOT Insight camera for bright-field microscopy; the images were assembled to movies using Photoshop CS5 and QuickTime Player Pro. For confocal live-imaging, image series were started as soon as the cellularization front had reached the yolk. Images were taken with 12 seconds/frame at a focal place of 5 µm below the apical surface.

Image quantification and statistical analysis

For automated cell tracking (Fig. 2) we developed custom scripts in MATLAB (MathWorks). Using low pass and high pass filtering, thresholding and built-in morphological operations, images were segmented to 1-pixel-thick cell outlines. Distances of cell centres between two subsequent images in the series were used as criterion to identify corresponding cells in the series. For each tracked cell pixel area and eccentricity was measured at each timepoint. Cell eccentricity was defined as the ratio of the major and minor axis of the ellipse that best fits the cell. For plotting, area and eccentricity time series were smoothened using a Gaussian kernel (σ = 1.5 time points) to reduce noise.

Apicobasal intensity profiles along membranes (Fig. 4) were measured using IPLab (Scanalytics) by manually drawing 1-pixel-thick linear ROIs along each membrane. For plotting, measurements were smoothened using a running average filter (window: +/− 10 data points).

For assessing signal diminishment of Dzy:GFP/Rap1:GFP (Fig. 5) or DE-Cad (Fig. 6) in the mesoderm, both mesodermal and ectodermal cells were marked manually in IPLab and mean pixel intensities were measured.

To quantify mesodermal spreading and clumping (Fig. 6), linear ROIs were drawn manually using IPLab according to the schemes in Fig. 6, and ROI lengths were measured.

To check for significance, R software (R Foundation for Statistical Computing) was used to apply nonparametric Wilcoxon rank sum tests.

Acknowledgements

We are indebted to A. Beermann and B. Moussian for critically reading the manuscript and to the Leptin lab for support during earlier phases of the project. In addition, we are grateful to B. Boettner, E. Wieschaus, N. Brown, J. Grosshans, E. Knust, L. Vogelsang, the Bloomington Stock Center and the DSHB for providing fly stocks or reagents. We thank M. Hoffmeister for providing Lamin/GFP co-stainings and S. Thürmann, E. Müller, U. Engels-Bochtler and K. Flessner for expert technical assistance.

Funding

This work was supported by the State Baden-Württemberg and the Reinhold-und-Maria-Teufel-Stiftung.

References

Asha
H.
,
de Ruiter
N. D.
,
Wang
M. G.
,
Hariharan
I. K.
(
1999
).
The Rap1 GTPase functions as a regulator of morphogenesis in vivo.
EMBO J.
18
,
605
615
.
Barrett
K.
,
Leptin
M.
,
Settleman
J.
(
1997
).
The Rho GTPase and a putative RhoGEF mediate a signaling pathway for the cell shape changes in Drosophila gastrulation.
Cell
91
,
905
915
.
Bilder
D.
(
2001
).
PDZ proteins and polarity: functions from the fly.
Trends Genet.
17
,
511
519
.
Boettner
B.
,
Harjes
P.
,
Ishimaru
S.
,
Heke
M.
,
Fan
H. Q.
,
Qin
Y.
,
Van Aelst
L.
,
Gaul
U.
(
2003
).
The AF-6 homolog canoe acts as a Rap1 effector during dorsal closure of the Drosophila embryo.
Genetics
165
,
159
169
.
Boettner
B.
,
Van Aelst
L.
(
2007
).
The Rap GTPase activator Drosophila PDZ-GEF regulates cell shape in epithelial migration and morphogenesis.
Mol. Cell. Biol.
27
,
7966
7980
.
Boettner
B.
,
Van Aelst
L.
(
2009
).
Control of cell adhesion dynamics by Rap1 signaling.
Curr. Opin. Cell Biol.
21
,
684
693
.
Brönner
G.
,
Chu–LaGraff
Q.
,
Doe
C. Q.
,
Cohen
B.
,
Weigel
D.
,
Taubert
H.
,
Jäckle
H.
(
1994
).
Sp1/egr-like zinc-finger protein required for endoderm specification and germ-layer formation in Drosophila.
Nature
369
,
664
668
.
Brown
N. H.
,
Kafatos
F. C.
(
1988
).
Functional cDNA libraries from Drosophila embryos.
J. Mol. Biol.
203
,
425
437
.
Campos–Ortega
J.
,
Hartenstein
V.
(
1985
).
The embryonic development of Drosophila melanogaster
.
Berlin, Heidelberg, New York
:
Springer-Verlag
.
Chou
T. B.
,
Perrimon
N.
(
1992
).
Use of a yeast site-specific recombinase to produce female germline chimeras in Drosophila.
Genetics
131
,
643
653
.
Chou
T. B.
,
Noll
E.
,
Perrimon
N.
(
1993
).
Autosomal P[ovoD1] dominant female-sterile insertions in Drosophila and their use in generating germ-line chimeras.
Development
119
,
1359
1369
.
Conte
V.
,
Muñoz
J. J.
,
Baum
B.
,
Miodownik
M.
(
2009
).
Robust mechanisms of ventral furrow invagination require the combination of cellular shape changes.
Phys. Biol.
6
,
016010
.
Dawes–Hoang
R. E.
,
Parmar
K. M.
,
Christiansen
A. E.
,
Phelps
C. B.
,
Brand
A. H.
,
Wieschaus
E. F.
(
2005
).
folded gastrulation, cell shape change and the control of myosin localization.
Development
132
,
4165
4178
.
de Beco
S.
,
Gueudry
C.
,
Amblard
F.
,
Coscoy
S.
(
2009
).
Endocytosis is required for E-cadherin redistribution at mature adherens junctions.
Proc. Natl. Acad. Sci. USA
106
,
7010
7015
.
de Rooij
J.
,
Boenink
N. M.
,
van Triest
M.
,
Cool
R. H.
,
Wittinghofer
A.
,
Bos
J. L.
(
1999
).
PDZ-GEF1, a guanine nucleotide exchange factor specific for Rap1 and Rap2.
J. Biol. Chem.
274
,
38125
38130
.
Dupuy
A. G.
,
L’Hoste
S.
,
Cherfils
J.
,
Camonis
J.
,
Gaudriault
G.
,
de Gunzburg
J.
(
2005
).
Novel Rap1 dominant-negative mutants interfere selectively with C3G and Epac.
Oncogene
24
,
4509
4520
.
Fox
D. T.
,
Homem
C. C.
,
Myster
S. H.
,
Wang
F.
,
Bain
E. E.
,
Peifer
M.
(
2005
).
Rho1 regulates Drosophila adherens junctions independently of p120ctn.
Development
132
,
4819
4831
.
Frasch
M.
,
Hoey
T.
,
Rushlow
C.
,
Doyle
H.
,
Levine
M.
(
1987
).
Characterization and localization of the even-skipped protein of Drosophila.
EMBO J.
6
,
749
759
.
Gates
J.
,
Peifer
M.
(
2005
).
Can 1000 reviews be wrong? Actin, alpha-Catenin, and adherens junctions.
Cell
123
,
769
772
.
Giebel
B.
,
Stüttem
I.
,
Hinz
U.
,
Campos–Ortega
J. A.
(
1997
).
Lethal of scute requires overexpression of daughterless to elicit ectopic neuronal development during embryogenesis in Drosophila.
Mech. Dev.
63
,
75
87
.
Grosshans
J.
,
Wenzl
C.
,
Herz
H. M.
,
Bartoszewski
S.
,
Schnorrer
F.
,
Vogt
N.
,
Schwarz
H.
,
Müller
H. A.
(
2005
).
RhoGEF2 and the formin Dia control the formation of the furrow canal by directed actin assembly during Drosophila cellularisation.
Development
132
,
1009
1020
.
Häcker
U.
,
Perrimon
N.
(
1998
).
DRhoGEF2 encodes a member of the Dbl family of oncogenes and controls cell shape changes during gastrulation in Drosophila.
Genes Dev
12
,
274
284
.
Harris
T. J. C.
,
Sawyer
J. K.
,
Peifer
M.
(
2009
).
How the cytoskeleton helps build the embryonic body plan: models of morphogenesis from Drosophila.
Curr. Top. Dev. Biol.
89
,
55
85
.
Haruta
T.
,
Warrior
R.
,
Yonemura
S.
,
Oda
H.
(
2010
).
The proximal half of the Drosophila E-cadherin extracellular region is dispensable for many cadherin-dependent events but required for ventral furrow formation.
Genes Cells
15
,
193
208
.
Hinz
U.
,
Giebel
B.
,
Campos–Ortega
J. A.
(
1994
).
The basic-helix-loop-helix domain of Drosophila lethal of scute protein is sufficient for proneural function and activates neurogenic genes.
Cell
76
,
77
87
.
Hogan
C.
,
Serpente
N.
,
Cogram
P.
,
Hosking
C. R.
,
Bialucha
C. U.
,
Feller
S. M.
,
Braga
V. M. M.
,
Birchmeier
W.
,
Fujita
Y.
(
2004
).
Rap1 regulates the formation of E-cadherin-based cell-cell contacts.
Mol. Cell. Biol.
24
,
6690
6700
.
Hoshino
T.
,
Sakisaka
T.
,
Baba
T.
,
Yamada
T.
,
Kimura
T.
,
Takai
Y.
(
2005
).
Regulation of E-cadherin endocytosis by nectin through afadin, Rap1, and p120ctn.
J. Biol. Chem.
280
,
24095
24103
.
Howard
S.
,
Deroo
T.
,
Fujita
Y.
,
Itasaki
N.
(
2011
).
A positive role of cadherin in Wnt/β-catenin signalling during epithelial-mesenchymal transition.
PLoS ONE
6
,
e23899
.
Huelsmann
S.
,
Hepper
C.
,
Marchese
D.
,
Knöll
C.
,
Reuter
R.
(
2006
).
The PDZ-GEF dizzy regulates cell shape of migrating macrophages via Rap1 and integrins in the Drosophila embryo.
Development
133
,
2915
2924
.
Hunter
C.
,
Wieschaus
E.
(
2000
).
Regulated expression of nullo is required for the formation of distinct apical and basal adherens junctions in the Drosophila blastoderm.
J. Cell Biol.
150
,
391
402
.
Ishimaru
S.
,
Williams
R.
,
Clark
E.
,
Hanafusa
H.
,
Gaul
U.
(
1999
).
Activation of the Drosophila C3G leads to cell fate changes and overproliferation during development, mediated by the RAS-MAPK pathway and RAP1.
EMBO J.
18
,
145
155
.
Kam
Z.
,
Minden
J. S.
,
Agard
D. A.
,
Sedat
J. W.
,
Leptin
M.
(
1991
).
Drosophila gastrulation: analysis of cell shape changes in living embryos by three-dimensional fluorescence microscopy.
Development
112
,
365
370
.
Kawajiri
A.
,
Itoh
N.
,
Fukata
M.
,
Nakagawa
M.
,
Yamaga
M.
,
Iwamatsu
A.
,
Kaibuchi
K.
(
2000
).
Identification of a novel β-catenin-interacting protein.
Biochem. Biophys. Res. Commun.
273
,
712
717
.
Kiehart
D. P.
,
Feghali
R.
(
1986
).
Cytoplasmic myosin from Drosophila melanogaster.
J. Cell Biol.
103
,
1517
1525
.
Knox
A. L.
,
Brown
N. H.
(
2002
).
Rap1 GTPase regulation of adherens junction positioning and cell adhesion.
Science
295
,
1285
1288
.
Kölsch
V.
,
Seher
T.
,
Fernandez–Ballester
G. J.
,
Serrano
L.
,
Leptin
M.
(
2007
).
Control of Drosophila gastrulation by apical localization of adherens junctions and RhoGEF2.
Science
315
,
384
386
.
Kuiperij
H. B.
,
de Rooij
J.
,
Rehmann
H.
,
van Triest
M.
,
Wittinghofer
A.
,
Bos
J. L.
,
Zwartkruis
F. J.
(
2003
).
Characterisation of PDZ-GEFs, a family of guanine nucleotide exchange factors specific for Rap1 and Rap2.
Biochim. Biophys. Acta
1593
,
141
149
.
Le
T. L.
,
Yap
A. S.
,
Stow
J. L.
(
1999
).
Recycling of E-cadherin: a potential mechanism for regulating cadherin dynamics.
J. Cell Biol.
146
,
219
232
.
Lee
J. H.
,
Cho
K. S.
,
Lee
J.
,
Kim
D.
,
Lee
S. B.
,
Yoo
J.
,
Cha
G. H.
,
Chung
J.
(
2002
).
Drosophila PDZ-GEF, a guanine nucleotide exchange factor for Rap1 GTPase, reveals a novel upstream regulatory mechanism in the mitogen-activated protein kinase signaling pathway.
Mol. Cell. Biol.
22
,
7658
7666
.
Leptin
M.
(
2005
).
Gastrulation movements: the logic and the nuts and bolts.
Dev. Cell
8
,
305
320
.
Leptin
M.
,
Grunewald
B.
(
1990
).
Cell shape changes during gastrulation in Drosophila.
Development
110
,
73
84
.
Martin
A. C.
,
Kaschube
M.
,
Wieschaus
E. F.
(
2009
).
Pulsed contractions of an actin-myosin network drive apical constriction.
Nature
457
,
495
499
.
Martin
A. C.
,
Gelbart
M.
,
Fernandez–Gonzalez
R.
,
Kaschube
M.
,
Wieschaus
E. F.
(
2010
).
Integration of contractile forces during tissue invagination.
J. Cell Biol.
188
,
735
749
.
McMahon
A.
,
Reeves
G. T.
,
Supatto
W.
,
Stathopoulos
A.
(
2010
).
Mesoderm migration in Drosophila is a multi-step process requiring FGF signaling and integrin activity.
Development
137
,
2167
2175
.
Mino
A.
,
Ohtsuka
T.
,
Inoue
E.
,
Takai
Y.
(
2000
).
Membrane-associated guanylate kinase with inverted orientation (MAGI)-1/brain angiogenesis inhibitor 1-associated protein (BAP1) as a scaffolding molecule for Rap small G protein GDP/GTP exchange protein at tight junctions.
Genes Cells
5
,
1009
1016
.
Mishra
S.
,
Smolik
S. M.
,
Forte
M. A.
,
Stork
P. J.
(
2005
).
Ras-independent activation of ERK signaling via the torso receptor tyrosine kinase is mediated by Rap1.
Curr. Biol.
15
,
366
370
.
Morin
X.
,
Daneman
R.
,
Zavortink
M.
,
Chia
W.
(
2001
).
A protein trap strategy to detect GFP-tagged proteins expressed from their endogenous loci in Drosophila.
Proc. Natl. Acad. Sci. USA
98
,
15050
15055
.
Morize
P.
,
Christiansen
A. E.
,
Costa
M.
,
Parks
S.
,
Wieschaus
E.
(
1998
).
Hyperactivation of the folded gastrulation pathway induces specific cell shape changes.
Development
125
,
589
597
.
Müller
H. A.
,
Wieschaus
E.
(
1996
).
armadillo, bazooka, and stardust are critical for early stages in formation of the zonula adherens and maintenance of the polarized blastoderm epithelium in Drosophila.
J. Cell Biol
134
,
149
163
.
Myster
S. H.
,
Cavallo
R.
,
Anderson
C. T.
,
Fox
D. T.
,
Peifer
M.
(
2003
).
Drosophila p120catenin plays a supporting role in cell adhesion but is not an essential adherens junction component.
J. Cell Biol
160
,
433
449
.
Nikolaidou
K. K.
,
Barrett
K.
(
2004
).
A Rho GTPase signaling pathway is used reiteratively in epithelial folding and potentially selects the outcome of Rho activation.
Curr. Biol.
14
,
1822
1826
.
Oda
H.
,
Tsukita
S.
(
2001
).
Real-time imaging of cell-cell adherens junctions reveals that Drosophila mesoderm invagination begins with two phases of apical constriction of cells.
J. Cell Sci.
114
,
493
501
.
Oda
H.
,
Tsukita
S.
,
Takeichi
M.
(
1998
).
Dynamic behavior of the cadherin-based cell-cell adhesion system during Drosophila gastrulation.
Dev. Biol.
203
,
435
450
.
Pacquelet
A.
,
Lin
L.
,
Rorth
P.
(
2003
).
Binding site for p120/delta-catenin is not required for Drosophila E-cadherin function in vivo.
J. Cell Biol.
160
,
313
319
.
Parks
S.
,
Wieschaus
E. F.
(
1991
).
The Drosophila gastrulation gene concertina encodes a G α-like protein.
Cell
64
,
447
458
.
Pham
N.
,
Cheglakov
I.
,
Koch
C. A.
,
de Hoog
C. L.
,
Moran
M. F.
,
Rotin
D.
(
2000
).
The guanine nucleotide exchange factor CNrasGEF activates ras in response to cAMP and cGMP.
Curr. Biol.
10
,
555
558
.
Ramain
P.
,
Heitzler
P.
,
Haenlin
M.
,
Simpson
P.
(
1993
).
pannier, a negative regulator of achaete and scute in Drosophila, encodes a zinc finger protein with homology to the vertebrate transcription factor GATA-1.
Development
119
,
1277
1291
.
Reuter
R.
,
Leptin
M.
(
1994
).
Interacting functions of snail, twist and huckebein during the early development of germ layers in Drosophila.
Development
120
,
1137
1150
.
Roth
S.
,
Stein
D.
,
Nüsslein–Volhard
C.
(
1989
).
A gradient of nuclear localization of the dorsal protein determines dorsoventral pattern in the Drosophila embryo.
Cell
59
,
1189
1202
.
Ryder
E.
,
Ashburner
M.
,
Bautista–Llacer
R.
,
Drummond
J.
,
Webster
J.
,
Johnson
G.
,
Morley
T.
,
Chan
Y. S.
,
Blows
F.
,
Coulson
D.
et al.  (
2007
).
The DrosDel deletion collection: a Drosophila genomewide chromosomal deficiency resource.
Genetics
177
,
615
629
.
Sawyer
J. K.
,
Harris
N. J.
,
Slep
K. C.
,
Gaul
U.
,
Peifer
M.
(
2009
).
The Drosophila afadin homologue Canoe regulates linkage of the actin cytoskeleton to adherens junctions during apical constriction.
J. Cell Biol.
186
,
57
73
.
Seher
T. C.
,
Narasimha
M.
,
Vogelsang
E.
,
Leptin
M.
(
2007
).
Analysis and reconstitution of the genetic cascade controlling early mesoderm morphogenesis in the Drosophila embryo.
Mech. Dev.
124
,
167
179
.
Seher
T. C.
,
Leptin
M.
(
2000
).
Tribbles, a cell-cycle brake that coordinates proliferation and morphogenesis during Drosophila gastrulation.
Curr. Biol.
10
,
623
629
.
Shirinian
M.
,
Popovic
M.
,
Grabbe
C.
,
Varshney
G.
,
Hugosson
F.
,
Bos
H.
,
Rehmann
H.
,
Palmer
R. H.
(
2010
).
The Rap1 guanine nucleotide exchange factor C3G is required for preservation of larval muscle integrity in Drosophila melanogaster.
PLoS ONE
5
,
e9403
.
Sweeton
D.
,
Parks
S.
,
Costa
M.
,
Wieschaus
E.
(
1991
).
Gastrulation in Drosophila: the formation of the ventral furrow and posterior midgut invaginations.
Development
112
,
775
789
.
Tautz
D.
,
Pfeifle
C.
(
1989
).
A non-radioactive in situ hybridization method for the localization of specific RNAs in Drosophila embryos reveals translational control of the segmentation gene hunchback.
Chromosoma
98
,
81
85
.
Tepass
U.
,
Gruszynski–DeFeo
E.
,
Haag
T. A.
,
Omatyar
L.
,
Török
T.
,
Hartenstein
V.
(
1996
).
shotgun encodes Drosophila E-cadherin and is preferentially required during cell rearrangement in the neurectoderm and other morphogenetically active epithelia.
Genes Dev.
10
,
672
685
.
Tepass
U.
,
Tanentzapf
G.
,
Ward
R.
,
Fehon
R. G.
(
2001
).
Epithelial cell polarity and cell junctions in Drosophila.
Annu. Rev. Genet.
35
,
747
784
.
Wang
H.
,
Singh
S. R.
,
Zheng
Z.
,
Oh
S. W.
,
Chen
X.
,
Edwards
K.
,
Hou
S. X.
(
2006
).
Rap-GEF signaling controls stem cell anchoring to their niche through regulating DE-cadherin-mediated cell adhesion in the Drosophila testis.
Dev. Cell
10
,
117
126
.
Young
P. E.
,
Pesacreta
T. C.
,
Kiehart
D. P.
(
1991
).
Dynamic changes in the distribution of cytoplasmic myosin during Drosophila embryogenesis.
Development
111
,
1
14
.

Supplementary information